arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2008.05248cs.LGstat.ML
收藏

可解释和公平表示的零抽样

Null-sampling for Interpretable and Fair Representations

University of Sussex(萨塞克斯大学)

Thomas Kehrenberg, Myles Bartlett, Oliver Thomas, Novi Quadrianto

摘要:

我们建议在数据域中学习不变表示,以实现算法公平性的可解释性。不变性意味着高水平的选择性,相关的相关性w.r.t.类别标签注释,以及对与受保护特征(诸如种族或性别)的不相关相关性的鲁棒性。我们引入了一个非平凡的设置,其中训练集表现出很强的偏见,使类标签注释是不相关的,虚假的相关性无法区分。为了解决这个问题,我们引入了一个逆向训练的模型与零采样过程中产生不变的表示在数据域。为了能够解开纠缠,使用部分标记的代表集。通过将表示放置到数据域中,模型所做的更改可以很容易地被人工审计员检查。我们展示了我们的方法在图像和表格数据集上的有效性:彩色MNIST,CelebA和成人数据集。

英文摘要:

We propose to learn invariant representations, in the data domain, to achieve interpretability in algorithmic fairness. Invariance implies a selectivity for high level, relevant correlations w.r.t. class label annotations, and a robustness to irrelevant correlations with protected characteristics such as race or gender. We introduce a non-trivial setup in which the training set exhibits a strong bias such that class label annotations are irrelevant and spurious correlations cannot be distinguished. To address this problem, we introduce an adversarially trained model with a null-sampling procedure to produce invariant representations in the data domain. To enable disentanglement, a partially-labelled representative set is used. By placing the representations into the data domain, the changes made by the model are easily examinable by human auditors. We show the effectiveness of our method on both image and tabular datasets: Coloured MNIST, the CelebA and the Adult dataset.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑